Enter your model's context window, system prompt size, tool schema size, reserved output headroom, and an average turn size. Get the fixed token overhead, what's left for conversation history, and how many turns fit before truncation risk — banded healthy, tight, or critical by how much of the window your fixed overhead alone consumes. Free.
Token counts are estimates you supply — from your provider's tokenizer, your framework's usage logs, or a rough word-count × 1.3 approximation. This tool does the arithmetic, not the counting.
Unlimited planning is free and stays free. The Optimization Kit adds what the arithmetic alone can't: concrete ways to shrink each fixed-overhead line item, and when to reach for summarization or retrieval instead of raw history.
Team licence for a platform/agents org? Email for a group rate.
Unlocks the Optimization Kit on this device and survives reloads. No account, nothing transmitted.